Return-Path: X-Original-To: apmail-flink-issues-archive@minotaur.apache.org Delivered-To: apmail-flink-issues-archive@minotaur.apache.org Received: from mail.apache.org (hermes.apache.org [140.211.11.3]) by minotaur.apache.org (Postfix) with SMTP id 32374186A2 for ; Wed, 13 Jan 2016 10:29:40 +0000 (UTC) Received: (qmail 35838 invoked by uid 500); 13 Jan 2016 10:29:40 -0000 Delivered-To: apmail-flink-issues-archive@flink.apache.org Received: (qmail 35775 invoked by uid 500); 13 Jan 2016 10:29:40 -0000 Mailing-List: contact issues-help@flink.apache.org; run by ezmlm Precedence: bulk List-Help: List-Unsubscribe: List-Post: List-Id: Reply-To: dev@flink.apache.org Delivered-To: mailing list issues@flink.apache.org Received: (qmail 35760 invoked by uid 99); 13 Jan 2016 10:29:40 -0000 Received: from arcas.apache.org (HELO arcas) (140.211.11.28) by apache.org (qpsmtpd/0.29) with ESMTP; Wed, 13 Jan 2016 10:29:40 +0000 Received: from arcas.apache.org (localhost [127.0.0.1]) by arcas (Postfix) with ESMTP id DFD262C0453 for ; Wed, 13 Jan 2016 10:29:39 +0000 (UTC) Date: Wed, 13 Jan 2016 10:29:39 +0000 (UTC) From: "Ufuk Celebi (JIRA)" To: issues@flink.apache.org Message-ID: In-Reply-To: References: Subject: [jira] [Commented] (FLINK-3151) YARN kills Flink TM containers due to memory overuse (outside heap/offheap) MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit X-JIRA-FingerPrint: 30527f35849b9dde25b450d4833f0394 [ https://issues.apache.org/jira/browse/FLINK-3151?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15095964#comment-15095964 ] Ufuk Celebi commented on FLINK-3151: ------------------------------------ I was able to reproduce this behaviour reliably on YARN with batch WordCount with a range partitioner and the following setup: 16 TMs with 1740 MB memory and a JM with 1024 MB and running with parallelism 16. > YARN kills Flink TM containers due to memory overuse (outside heap/offheap) > --------------------------------------------------------------------------- > > Key: FLINK-3151 > URL: https://issues.apache.org/jira/browse/FLINK-3151 > Project: Flink > Issue Type: Bug > Components: TaskManager > Affects Versions: 1.0.0, 0.10.1 > Environment: 48 cores > Reporter: Robert Metzger > Assignee: Ufuk Celebi > Priority: Blocker > Fix For: 1.0.0, 0.10.2 > > > A Flink user who's running Flink on YARN with 1 processing slot, 2 GB of TM memory on a machine with 48 reported CPU cores is running into issues with TM containers being killed due to memory overuse. > Setting the YARN memory cutoff to 0.5 resolves the problem, but its not really a feasible approach. > Another solution to the problem was downgrading netty again from 4.0.31.Final to 4.0.27.Final resolved the issue. > We upgraded Netty between 0.9 and 0.10. > Most likely netty changed its behavior between the releases. -- This message was sent by Atlassian JIRA (v6.3.4#6332)